Numeric Simulation Laboratory - Davide Nicoli

My solutions to the assignments given in the 2021/2022 edition of the laboratory.

Repository structure

The root directory of the project is organized in 5 directories plus some support files:

• data/: where configurations and dataset files are stored, for example the prime numbers lists for the random generator or the molecular dynamics settings;
• results/: where results from simulations and samplings are stored by default. Each exercise has its own subdirectory; some results are already cached, others will be computed on the fly;
• solutions/: it contains jupyter notebooks presenting the exercises' solutions;
• src/: it contains the C++ code

How to display the exercises results

Launch a jupyter server from the solutions/ directory:

cd solutions
jupyter notebook

The jupyter notebooks that display the solutions to the exercises are stored under solutions/. Provided that a jupyter server is available to the user, every notebook should be ready for consultation.

If a full recomputation of results is wanted instead, the following section presents which steps need to be taken in order to prepare the source tree and install python runtime dependencies. Running the notebooks then triggers recomputation, which may take a while. Many notebooks provide the option to use cached results setting the variable USE_CACHED to True: errors related to files not being found are usually related to the fact that solutions are not in cache. Try setting USE_CACHED=False if that is the case.

How to prepare the project for recomputation

Project dependencies

The following software needs to be present on the computer in order to build the project:

• a modern compiler supporting c++17 (the project was tested with gcc-10, gcc-12 and clang-14);
• cmake >= 3.18;
• an anaconda distribution;
• an MPI distribution (both mpich and OpenMPI were tested).

Build and runtime dependencies, that will instead be installed in the following sections, are specified in:

• "environment.yml" (for python dependencies; managed by anaconda);
• "conanfile.txt" (for C++ dependencies; managed by conan, which will be installed by conan).

How to build the project

After having installed the required build dependencies and cloned the current repository to a folder, the following steps should be followed:

1. Create a conda environment and install python dependencies

From the project root run:

conda env create -f environment.yml -n <YourEnvName>

This command creates a conda environment named <YourEnvName> and installs the dependencies listed in "environment.yml". Now activate your environment:

conda activate <YourEnvName>

2. Configure Conan and install C++ dependencies

C++ dependencies are managed by conan. It was installed by conda during the previous step, and now it has to be configured:

conan profile new default --detect

the prompt may show a warning message starting with:

************************* WARNING: GCC OLD ABI COMPATIBILITY ***********************

In case, I suggest following the directions given in the warning and execute:

conan profile update settings.compiler.libcxx=libstdc++11 default

to make the software work properly with code compiled for recent versions of C++.

Dependencies can now be installed. Start with creating a build directory:

mkdir build
cd build

where the cmake configuration files and the build tree will reside. Now install C++ dependencies running:

conan install -pr:h=default -pr:b=default --build=missing -s compiler.cppstd=17 ..

3. Build the project using cmake

Inside the build/ directory created above run:

cmake .. -DCMAKE_BUILD_TYPE=Release -DCMAKE_INSTALL_PREFIX=../
cmake --build .
cmake --install .

This should build and install the executables used to solve the exercises under the bin/ directory in the project's root, ready to be used by the jupyter notebooks.